AI-based vehicle paint inspection processing system

Abstract

The application discloses an AI-based vehicle paint surface detection processing system and belongs to the technical field of vehicle paint surface detection. The system comprises a platform end and a user end. The platform end comprises a paint surface detection module and a detection guidance module. The paint surface detection module is used for detecting the paint surface of a vehicle to be detected, obtaining vehicle information and a paint surface image of the vehicle to be detected, analyzing the vehicle information and the paint surface image, obtaining a paint surface detection result, and sending the paint surface detection result to a corresponding user end. The detection guidance module is used for setting a paint surface collection tutorial of a corresponding vehicle, and storing the paint surface collection tutorial in a tutorial library. The user end comprises a paint surface collection module and a result display module. The paint surface collection module is used for collecting a paint surface image of a user's vehicle, and sending vehicle information and the paint surface image to the platform end. The result display module is used for displaying a paint surface detection result, generating a vehicle model according to the vehicle information, obtaining the paint surface detection result in real time, and displaying and adjusting the vehicle model according to the paint surface detection result.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of vehicle paint detection, and particularly relates to an AI-based vehicle paint detection processing system. BACKGROUND

[0002] In the daily use of personal vehicles and vehicle transaction scenarios, the paint quality directly affects the vehicle value, use experience and transaction rights and interests, so how to detect the vehicle paint has become a problem that personal vehicle owners urgently need to solve. However, the current ordinary vehicle owners often rely on non-professional channels for paint detection, such as roadside repair shops, non-chain beauty shops, etc., and the detection means mainly rely on manual experience judgment, lack of unified standards, resulting in high error rate in identifying defects such as scratches, color difference and orange peel lines, and low credibility. If personal vehicle owners need accurate detection, they need to entrust third-party institutions such as 4S shops and professional detection companies, which is high in single detection cost and needs to be reserved and waited, and the process is complicated.

[0003] Therefore, in order to meet the efficient needs of personal vehicle owners for vehicle paint detection, the application provides an AI-based vehicle paint detection processing system. SUMMARY

[0004] In order to solve the problems existing in the above scheme, the application provides an AI-based vehicle paint detection processing system.

[0005] The purpose of the application can be achieved by the following technical scheme:

[0006] The AI-based vehicle paint detection processing system comprises a platform end and a user end.

[0007] The platform end comprises a paint detection module and a detection guidance module.

[0008] The paint detection module is used for paint detection of a vehicle to be detected, obtains vehicle information and a paint image of the vehicle to be detected, analyzes the vehicle information and the paint image, obtains a paint detection result, and sends the paint detection result to the corresponding user end.

[0009] The detection guidance module is used for setting a paint collection tutorial of the corresponding vehicle, marking the vehicle information of the paint collection tutorial, and storing the paint collection tutorial in a tutorial library.

[0010] Further, the setting of the paint collection tutorial comprises:

[0011] A vehicle detail table is set, which is used for counting various vehicle information, and a paint feature set of the corresponding vehicle is collected in real time according to the vehicle detail table.

[0012] determine whether the corresponding vehicles meet the tutorial use standard according to the paint feature set, the tutorial use standard being able to collect paint images using the same paint collection tutorial;

[0013] vehicles meeting the tutorial use standard are classified into a category, and a corresponding vehicle paint classification is obtained; the platform party sets a corresponding paint collection tutorial for the vehicle paint classification.

[0014] Further, when a new vehicle paint classification is obtained, the paint feature set of the corresponding vehicle is identified, and a vehicle paint classification with the highest similarity is matched according to the paint feature set; the paint collection tutorial of the vehicle paint classification is identified, and the paint collection tutorial is adjusted according to the paint feature set, to obtain a paint collection tutorial of the new vehicle paint classification.

[0015] The user end includes a paint collection module and a result display module.

[0016] The paint collection module is used to collect paint images of the user's vehicle, and send vehicle information and paint images to the platform end.

[0017] Further, the paint image collection method includes:

[0018] Obtain vehicle information of the user's vehicle, and match a corresponding paint collection tutorial from a tutorial library according to the vehicle information; and display the paint collection tutorial to the user.

[0019] The user collects images of the vehicle paint according to the paint collection tutorial, to obtain paint images.

[0020] Further, the paint collection module further includes a docking collection unit, which is used to dock an image collection system of the user's vehicle, and the image collection system is used to collect images of the user's vehicle.

[0021] Real-time collection of vehicle collection images of the user's vehicle by the image collection system, processing of the vehicle collection images, and obtaining of vehicle paint images.

[0022] Further, the processing of the collection images includes:

[0023] Positioning of the collection images according to positions of the user's vehicle, quality calibration of the collection images of the corresponding positions of the user's vehicle, obtaining of quality calibration results of the collection images of the corresponding positions, the quality calibration results including quality calibration qualified and quality calibration unqualified, and corresponding processing of the collection images according to the quality calibration results.

[0024] Further, the quality calibration of the collection images of the corresponding positions of the user's vehicle includes:

[0025] A quality calibration model is established, and an expression of the quality calibration model is:

[0026]

[0027] In the formula, s is input data, indicating a collected image; and output data is a quality evaluation value ZP(s), and the quality evaluation value is 1 or 0;

[0028] The collected image of the corresponding position is analyzed by the quality calibration model to obtain a quality evaluation value of the corresponding position;

[0029] When the quality evaluation value is 1, the quality calibration result is quality calibration qualified;

[0030] When the quality evaluation value is 0, the quality calibration result is quality calibration unqualified.

[0031] Further, an image quality model is generated according to the quality calibration results of each position of the user's vehicle obtained in real time, and the image quality model is displayed to the user in real time.

[0032] Further, the image quality model is generated as follows:

[0033] A visual model of the user's vehicle is generated according to vehicle information, and is marked as a vehicle model; a display mode corresponding to quality calibration qualified and quality calibration unqualified is set respectively;

[0034] The quality calibration result of the corresponding position of the vehicle is obtained in real time, the paint surface area in the vehicle model is displayed and adjusted according to the display mode corresponding to the quality calibration result, and an image quality model is obtained; and the image quality model is dynamically updated according to the quality calibration result.

[0035] The result display module is used to display the paint surface detection result, a vehicle model is generated according to vehicle information, the paint surface detection result is obtained in real time, the vehicle model is displayed and adjusted according to the paint surface detection result; and the vehicle model after the display adjustment is displayed to the user.

[0036] Further, the vehicle model is updated in real time according to the paint surface detection result.

[0037] Further, the vehicle model is displayed and adjusted according to the paint surface detection result, including:

[0038] A display adjustment mode of the vehicle model corresponding to different paint surface detection results is preset; the paint surface detection result is identified in real time, the corresponding display adjustment mode is matched according to the paint surface detection result, and the vehicle model is displayed and adjusted according to the display adjustment mode;

[0039] When the paint surface detection result is detection abnormality, a paint surface image of the corresponding vehicle paint surface is obtained; and the paint surface image and the paint surface detection result are supplemented into the vehicle model.

[0040] Compared with the prior art, the beneficial effects of the present application are:

[0041] Through the cooperation between the modules, the expensive equipment and manual service of the 4S store / professional testing agency are replaced by the AI algorithm, the single detection cost is reduced, the individual car owner can obtain a professional level detection report without paying high fees, the real-time paint surface image of the user's vehicle is collected through the docking collection unit, the real-time paint surface detection of the user's vehicle is realized, the user is helped to find the paint surface anomaly in time, and for the user who does not match the paint surface collection tutorial, the intelligent adjustment mode is adopted, so that the user can collect the paint surface image, and long waiting time of the user is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only represent some embodiments of the present application, and all other drawings obtained by those skilled in the art without creative labor on the premise of the drawings also belong to the protection scope of the present application.

[0043] Figure 1 The present application is a schematic diagram of the principle. DETAILED DESCRIPTION

[0044] The technical solutions of the present application will be described in detail below in combination with the embodiments. Obviously, the described embodiments only represent some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor on the premise of the drawings also belong to the protection scope of the present application.

[0045] As shown in Figure 1 the vehicle paint surface detection processing system based on AI includes a platform end and a user end;

[0046] The platform end is maintained and used by the platform party, including a paint surface detection module and a detection guidance module.

[0047] The paint surface detection module is used for paint surface detection of the vehicle, obtaining vehicle information and paint surface images of the vehicle to be detected; analyzing the vehicle information and the paint surface images to obtain a paint surface detection result; and sending the paint surface detection result to the corresponding user end.

[0048] In one embodiment, the vehicle information and the paint surface images are analyzed, and the platform party analyzes according to the existing AI technology and paint surface detection technology, such as establishing a corresponding intelligent detection model for intelligent detection.

[0049] For example, based on the lightweight YOLOv8-Tiny model, common defects such as scratches and sagging are located; the suspicious area calls the Transformer-3D model, and the defect depth, area and repair cost are calculated combined with multi-modal data; for user vehicles, the model is fine-tuned with a small number of samples; an interactive 3D model is generated through WebGL, and users can rotate to view the defect location and simulate the effect after repair.

[0050] The detection guidance module is configured to set a paint surface collection tutorial for a corresponding vehicle, mark the paint surface collection tutorial with corresponding vehicle information, and store the paint surface collection tutorial in a tutorial library. The tutorial library is a database for storing data.

[0051] In one embodiment, the platform party can set the paint surface collection tutorial for the corresponding vehicle in the existing manner.

[0052] In one embodiment, the setting of the paint surface collection tutorial includes:

[0053] In one embodiment, the setting of the paint surface collection tutorial includes:

[0054] In one embodiment, the setting of the paint surface collection tutorial includes:

[0055] In one embodiment, the setting of the paint surface collection tutorial includes:

[0056] In one embodiment, when a new vehicle paint surface classification is obtained, the paint surface feature set corresponding to the vehicle is identified, and a vehicle paint surface classification with the highest similarity is matched according to the paint surface feature set. The paint surface collection tutorial corresponding to the vehicle paint surface classification is identified, and the paint surface collection tutorial is adjusted according to the paint surface feature set of the vehicle to obtain a new paint surface collection tutorial for the vehicle paint surface classification. Manual adjustment can be used, or intelligent adjustment based on existing intelligent technology can be used.

[0057] For the user matching without the paint surface collection tutorial, the intelligent adjustment mode is adopted, so as to facilitate the user to collect the paint surface image and avoid the long waiting of the user.

[0058] The user end is used by the user and can be set in the form of a small program, an app, a webpage and the like; and includes a paint surface collection module and a result display module.

[0059] The paint surface collection module is used to collect the paint surface image of the vehicle of the user, acquire the vehicle information of the vehicle of the user, and send the vehicle information and the paint surface image to the platform end.

[0060] In one embodiment, the collection method of the paint surface image includes:

[0061] Acquiring the vehicle information of the vehicle of the user, matching the corresponding paint surface collection tutorial from the tutorial library according to the vehicle information, and displaying the paint surface collection tutorial to the user.

[0062] The user collects the image of the vehicle paint surface according to the paint surface collection tutorial, and obtains the paint surface image.

[0063] In one embodiment, the paint surface image collected by the user can also be checked to guarantee the quality of the paint surface image.

[0064] In one embodiment, the paint surface collection module further includes a docking collection unit, which is used to dock the image collection system of the vehicle panoramic camera system, 360° ring view monitoring system and the like of the vehicle of the user, acquire the vehicle collection image of the vehicle of the user in real time, process the vehicle collection image, and obtain the vehicle paint surface image.

[0065] The real-time paint surface image collection of the vehicle of the user is realized through the docking collection unit, the real-time paint surface detection of the vehicle of the user is realized, and the user is helped to discover the paint surface anomaly in time.

[0066] In one embodiment, the processing of the collection image can adopt the existing image processing technology to crop the vehicle collection image, and obtain the paint surface image of the vehicle of the user.

[0067] In one embodiment, the processing of the collection image includes:

[0068] The collected images are positioned according to the user vehicle position, the collected images of the corresponding position of the user vehicle are quality calibrated, the quality calibration result of the collected images of the corresponding position is obtained, including quality calibration qualified and quality calibration unqualified; that is, both the image quality and the excessive pollutants are regarded as quality calibration unqualified, the collected images of quality calibration unqualified will affect the subsequent detection analysis, therefore, the subsequent elimination is carried out, because it is dynamic real-time collection and detection, therefore, there is a time span, after the current image is eliminated, the detection is carried out according to the subsequent images meeting the requirements, the interval is short, and the detection precision is improved; the collected images are processed according to the quality calibration result, such as marking unnecessary detection analysis, cropping and eliminating and the like; and the backward user can be recorded for feedback.

[0069] In one embodiment, the collected images of the corresponding position of the user vehicle are quality calibrated, which can be calibrated based on the existing image quality calibration mode.

[0070] In one embodiment, the collected images of the corresponding position of the user vehicle are quality calibrated, including:

[0071] A quality calibration model is established, and the expression of the quality calibration model is:

[0072]

[0073] In the formula, s is input data, indicating the collected images; the historical image marking training set is used for training, the characteristics of various pollutants possibly possessed by the vehicle and the characteristics of the image quality not meeting the requirements are determined, the training set is set by summarizing, so that the subsequent collected images of the corresponding position can be judged whether they meet the image quality requirements; the output data is a quality evaluation value ZP(s), and the quality evaluation value is 1 or 0;

[0074] The collected images of the corresponding position are analyzed by the quality calibration model, and the quality evaluation value of the corresponding position is obtained.

[0075] When the quality evaluation value is 1, the quality calibration result is quality calibration qualified.

[0076] When the quality evaluation value is 0, the quality calibration result is quality calibration unqualified.

[0077] In one embodiment, the corresponding image quality model is generated according to the quality calibration results of each position of the user vehicle obtained in real time, and the image quality model is displayed to the user in real time.

[0078] The image quality model is generated according to the vehicle paint surface, is a visual data model, and can be a two-dimensional or three-dimensional model.

[0079] The generation of the image quality model:

[0080] According to the vehicle information, a visual model of the user's vehicle is generated, which is marked as a vehicle model; and a quality calibration qualified and a quality calibration unqualified are respectively set to correspond to display modes, such as green, red, and the like;

[0081] The quality calibration result of the corresponding position of the vehicle is acquired in real time, the paint surface area in the vehicle model is displayed and adjusted according to the display mode corresponding to the quality calibration result, the paint surface area is the area corresponding to the paint surface of the vehicle, the vehicle model after the display adjustment is marked as an image quality model; and the image quality model is dynamically updated according to the quality calibration result.

[0082] In one embodiment, the user can also be recommended to take corresponding optimization measures, such as replacing a device with higher image acquisition capability, repairing and adjusting through an image repair algorithm, and the like, according to the image quality model.

[0083] The result display module is configured to display the paint surface detection result, generate a vehicle model according to vehicle information, acquire the paint surface detection result sent by a platform party in real time, display and adjust the vehicle model according to the paint surface detection result; display the vehicle model after the display adjustment to the user, and update the vehicle model in real time according to the paint surface detection result.

[0084] The vehicle model is displayed and adjusted according to the paint surface detection result, the paint surface detection unqualified is displayed differently, and the corresponding paint surface detection result is supplemented, so as to facilitate the user to view.

[0085] For example, the vehicle model is displayed and adjusted according to the paint surface detection result, which includes:

[0086] The display adjustment mode of the vehicle model corresponding to different paint surface detection results is preset, a plurality of display adjustment modes can be preset, and the display adjustment mode corresponding to different paint surface detection results can be selected according to the user's preference;

[0087] The paint surface detection result is identified in real time, the corresponding display adjustment mode is matched according to the paint surface detection result, and the vehicle model is displayed and adjusted according to the display adjustment mode.

[0088] In one embodiment, when the paint surface detection result is detection abnormality, the paint surface image of the corresponding paint surface is acquired, the paint surface image and the paint surface detection result are supplemented to the vehicle model, so as to facilitate the user to check, and the paint surface detection technology is optimized and adjusted according to the user's checking result.

[0089] The above formulas are all calculated by removing the dimension and taking the numerical value, the formula is obtained by software simulation of a large amount of data to obtain a formula closest to the real situation, and the preset parameters and the preset threshold in the formula are set by a person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0090] The above examples are only used to illustrate the technical method of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present application.

Claims

1. An AI-based vehicle paint detection processing system, characterized by, The platform end and the user end are included; The platform end includes a paint surface detection module and a detection guidance module; The paint surface detection module is used for paint surface detection on a vehicle to be detected, obtaining vehicle information and a paint surface image of the vehicle to be detected; analyzing the vehicle information and the paint surface image to obtain a paint surface detection result; and sending the paint surface detection result to a corresponding user end; The detection guidance module is used for setting a paint surface collection tutorial for a corresponding vehicle, marking corresponding vehicle information for the paint surface collection tutorial, and storing the paint surface collection tutorial in a tutorial library; The user end includes a paint surface collection module and a result display module; The paint surface collection module is used for collecting a paint surface image of a user vehicle and sending the vehicle information and the paint surface image to the platform end; The result display module is used for paint surface detection result display, generating a vehicle model according to the vehicle information, obtaining a paint surface detection result in real time, displaying and adjusting the vehicle model according to the paint surface detection result, and displaying the vehicle model after the display adjustment to the user; The paint surface collection module further includes a docking collection unit, which is used for docking an image collection system of the user vehicle, and the image collection system is used for image collection on the user vehicle; The vehicle collection image of the user vehicle collected by the image collection system is obtained in real time, and the vehicle collection image is processed to obtain a vehicle paint surface image; Processing the collection image includes: Positioning the collection image according to the position of the user vehicle, quality calibration on the collection image of the corresponding position of the user vehicle, obtaining a quality calibration result of the collection image of the corresponding position, and the quality calibration result includes quality calibration qualified and quality calibration unqualified; and processing the collection image according to the quality calibration result; Generating an image quality model according to the quality calibration result of each position of the user vehicle obtained in real time, and displaying the image quality model to the user in real time; Generation of the image quality model: Generating a visual model of the user vehicle according to the vehicle information, and marking the visual model as a vehicle model; and setting a display mode corresponding to quality calibration qualified and quality calibration unqualified, respectively; Real-time acquisition of the quality calibration result of the corresponding position of the vehicle, display adjustment on the paint surface area in the vehicle model according to the display mode corresponding to the quality calibration result, and obtaining the image quality model; and dynamic updating of the image quality model according to the quality calibration result. 2.The AI-based vehicle paint detection processing system of claim 1, wherein Setting of the paint surface collection tutorial includes: Setting a vehicle detail table, which is used for statistical various vehicle information; and real-time collection of a paint surface feature set of a corresponding vehicle according to the vehicle detail table; Determining whether the corresponding vehicles meet a tutorial use standard according to the paint surface feature set, the tutorial use standard being that the paint surface images can be collected by using the same paint surface collection tutorial; Classifying the vehicle information meeting the tutorial use standard into a category, obtaining a corresponding vehicle paint surface classification, and setting a corresponding paint surface collection tutorial for the vehicle paint surface classification by the platform. 3.The AI-based vehicle paint detection processing system of claim 2, wherein When a new vehicle paint classification is obtained, a paint feature set of the corresponding vehicle is identified, and a vehicle paint classification with the highest similarity is matched according to the paint feature set; a paint collection tutorial of the vehicle paint classification is identified, the paint collection tutorial is adjusted according to the paint feature set, and a paint collection tutorial of the new vehicle paint classification is obtained. 4.The AI-based vehicle paint detection processing system of claim 1, wherein The paint image collection method comprises: Obtaining vehicle information of a user's vehicle, and matching a corresponding paint collection tutorial from a tutorial library according to the vehicle information; and displaying the paint collection tutorial to the user; The user collects images of the vehicle paint according to the paint collection tutorial, and obtains paint images. 5.The AI-based vehicle paint detection processing system of claim 1, wherein Quality calibration of the collected images of the corresponding positions of the user's vehicle comprises: A quality calibration model is established, and an expression of the quality calibration model is: ; In the formula, s is input data representing the collected images; and the output data is a quality evaluation value TP(s), which is 1 or 0; The collected images of the corresponding positions are analyzed by the quality calibration model to obtain the quality evaluation value of the corresponding positions; When the quality evaluation value is 1, the quality calibration result is quality calibration qualified; When the quality evaluation value is 0, the quality calibration result is quality calibration unqualified. 6.The AI-based vehicle paint detection processing system of claim 1, wherein Display adjustment of the vehicle model according to the paint detection result comprises: A display adjustment mode of the vehicle model according to different paint detection results is preset; the paint detection result is identified in real time, the corresponding display adjustment mode is matched according to the paint detection result, and the display adjustment of the vehicle model is performed according to the display adjustment mode; When the paint detection result is abnormal detection, the paint image of the corresponding vehicle paint is obtained; and the paint image and the paint detection result are supplemented into the vehicle model.

Images